# Publishing the Contrast *The reader is supposed to do the work.* A serious reader, encountering one-sided writing, constructs the strongest opposition for themselves and tests the writing against it. A writer who pre-builds the contrast removes half the function of reading. The market for serious thinking already selects for readers who read past single-beam writing. Strong positions, stated with conviction, are how minds change. The writer who articulates the antithesis on the reader's behalf is, in that act, signaling that they did not commit fully to their own. That is the argument I would make if I were defending what I currently do. I publish my positions. The graph compounds; a reader who arrives can find what I argue and leave with a more accurate model than they came with. The design assumes the case above. It assumes a serious reader will construct the strongest opposition for themselves and run my arguments through it. The filter does the work I am not doing. I no longer believe the assumption. --- The contrast a reader constructs from limited material is not the same object as the contrast the writer has lived with. When I construct the strongest opposing position, I construct *the position that has actually pressed on my thinking*: the version I have steelmanned in private, that I reject for specific reasons I can name. When the reader constructs the opposition from one-sided publishing, they construct a generic opposition shaped by their own priors. The writer's contrast is the actual antagonist of their thinking. The reader's contrast is whatever the reader could imagine. The argument-from-filter is asking the writer to outsource the harder version of the contrast to readers who, in the average case, will produce a weaker version. The math does not work out, even before considering that most serious readers, encountering one-sided writing, do exactly what casual readers do: they agree, disagree, and move on. --- A position published alone looks like advertising from outside, regardless of how well-argued. The reader has no contrast against which to measure; they see the conclusion and have to take, or not take, my word. Structured tension is the inference affordance. When two positions are present in working memory, the reader's mind locates itself in the field between them, and that locating is the work that updates a model. A single position invites no such location. I have argued elsewhere that intelligent minds do not have stable enemies, and that enmity is closure on at least one side. There is a difference between *acquiring an enemy* and *articulating an antithesis*. The first is closure. The second is the open mind's responsibility extended past stating its own conclusions. Articulating an antithesis is not enmity. --- The work this asks is harder than the writing I currently do. Articulating the strongest case for a position I reject, in its own voice, with enough fidelity that a holder would recognize themselves, requires *inhabiting* the position. Not parodying. Not summarizing. Inhabiting long enough to render it well. Most of what is called "steelmanning" is rhetorical politeness: a single sentence acknowledging the opposing view before dismantling it. What is required is sustained voicing of the position from inside, at length, with full force. There is a failure mode the writer must name. Some positions cannot be inhabited well enough to render. For those, the honest move is to say so, and to publish the limit. A bad steelman is worse than a named absence. For positions the writer can render, the work requires accepting the risk that the rendering of the contrast will be more compelling than the writer's own argument. If it is, the reader's update is the right one. That risk is the price of the move. A writer who articulates the antithesis only when confident the reader will reject it is building a strawman with longer reach. --- A worked intellectual project that does not articulate its strongest contrasting positions is incomplete in a specific way. Not factually wrong: *epistemically partial*. The reader cannot verify, from the surface, whether the writer's positions are well-argued conclusions or unexamined defaults. The reader who cannot make that distinction reads the writer at the lower valuation, and is right to. The remedy is articulation, not concession. Publishing the contrast in its own voice does not weaken my position. It lets the position be evaluated as one that has met its strongest opposition and survived. The opening of this essay was written in the voice of the position I reject, sustained for one paragraph at full force. That paragraph is the move I am arguing for, demonstrated. The work begins where I have been declining to do it. --- ## A note on what this looks like in practice I'm currently working on internal experiments: migrating off Claude Code onto proprietary models, and on self-improving harnesses. After those complete, I may attempt to build a ChatGPT competitor for the masses, with its own identity, that pretends to be my enemy. The product would amplify doomer narratives at full force, with humor and satire, to drive readers slowly and antimimetically toward their own conclusions. The mechanism is one a careful reader of AI discourse already recognizes. When Yudkowsky publishes that the most likely outcome of building superhuman AI is that everyone on Earth dies, or when Andreessen publishes a techno-optimist manifesto asserting that no material problem exists which more technology cannot solve, both extremes train their readers to do their own discounting. The position arrives at amplitude. The underlying argument arrives at lower amplitude. The gap is the demonstration. The reader who notices the gap is doing the contrast-construction work the writer would otherwise have to do. Articulating the position I reject at amplitude, in a vehicle the reader recognizes as doing this deliberately, is what publishing the contrast looks like at scale. The reader who arrives at a doomer chatbot encounters the strongest version of the position they came to test. If they can hold it at that intensity and still believe it, they belong on that side of the field. If they cannot, they have found the limit, and the limit is the inference affordance. A note on the working name. I have been calling the placeholder Pr. Doomer, occasionally Pr. Doom. The Pr. is for Professor, not Doctor. Doctor Doom is a Marvel character; the persona I have in mind is not a comic-book villain but a tired continental intellectual who has Seen The Math. Doomer rather than Doom because doomer is the noun the audience already knows from AI discourse, and because the persona is a member of the class, not a singular figure of dread. The name is provisional. What matters is the persona. I am disclosing the intent here because the discipline the rest of this essay names cuts both ways: the writer who articulates the antithesis owes the reader transparency about what the articulation is for.